Communication network traffic prediction method and apparatus

Abstract

The application discloses a communication network traffic prediction method and device, and relates to the field of time series prediction. The communication network traffic prediction method comprises the following steps: acquiring cell traffic data; determining a first traffic threshold according to the cell traffic data; determining an adjacency matrix in graph convolution of a traffic prediction model; inputting the cell traffic data, the first traffic threshold and the adjacency matrix into the traffic prediction model to acquire a prediction result. The above scheme can improve the accuracy of traffic prediction.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of time series prediction, and particularly relates to a communication network traffic prediction method and device. BACKGROUND

[0002] Communication network traffic prediction data has obvious heterogeneity, and different cell traffic data characteristics have great differences. The data is periodic and the trend changes are extremely irregular, and it is very challenging to predict cell traffic data. The existing prediction methods are mostly for traffic flow prediction, and the effect of directly using them for communication network traffic prediction is not good. In order to support the prediction of communication traffic and other complex data, the existing mainstream solutions are as follows.

[0003] The first solution is to model time and space separately, and use common graph networks to extract spatial features and use convolutional neural networks to extract temporal features. Graph networks require adjacency matrix information between nodes, and in the communication traffic scenario, the adjacency matrix is generally constructed according to the distance or correlation between nodes, but some prior information such as cell type is also helpful for spatial relationship.

[0004] The second solution is to derive features from data and then use machine learning algorithms for training and prediction.

[0005] The existing communication network traffic prediction technology has the following shortcomings:

[0006] 1. Low accuracy, which does not meet the requirements of actual application.

[0007] 2. Insufficient expression of spatial relationship between cells.

[0008] 3. No follow-up application related to prediction results. SUMMARY

[0009] The embodiments of the application provide a communication network traffic prediction method and device, which can solve the problem of low accuracy of the existing communication network traffic prediction method.

[0010] In order to solve the above technical problems, the embodiments of the application provide a communication network traffic prediction method, characterized in that it comprises:

[0011] Obtaining cell traffic data;

[0012] Determining a first traffic threshold according to the cell traffic data;

[0013] Determining an adjacency matrix in graph convolution of a traffic prediction model;

[0014] Inputting the cell traffic data, the first traffic threshold and the adjacency matrix into the traffic prediction model to obtain a prediction result.

[0015] Optionally, the determining the first traffic threshold according to the cell traffic data comprises:

[0016] grouping the cell traffic data according to time points, wherein the traffic data of different cells belonging to the same time point belong to the same group;

[0017] determining the traffic value corresponding to the traffic data located at the preset percentile point in each group as the first traffic threshold.

[0018] Optionally, the determining the adjacency matrix in the graph convolution of the traffic prediction model comprises:

[0019] determining the adjacency matrix in the graph convolution of the traffic prediction model based on the cell type and the inter-cell switching frequency.

[0020] Optionally, the determining the adjacency matrix in the graph convolution of the traffic prediction model based on the cell type and the inter-cell switching frequency comprises:

[0021] according to the formula: determining the adjacency matrix in the graph convolution of the traffic prediction model;

[0022] wherein A is the adjacency matrix in the graph convolution of the traffic prediction model, A i,j is the value corresponding to the i-th row and j-th column of the adjacency matrix, dis(i,j) is the distance between cell i and cell j, d_thres is a pre-set neighbor distance threshold, Type i is the type of cell i, Type j is the type of cell j, and switch_freq(i,j) is the inter-cell switching frequency, count i,j is the number of switching between cell i and cell j, and total_time is the number of time points of data.

[0023] Optionally, the loss function in the traffic prediction model is determined based on the first traffic threshold.

[0024] Optionally, the loss function is obtained in the following manner:

[0025] according to the formula:

[0026] determining the loss function;

[0027] wherein, is the loss function; Ωupper is a set of time points in the traffic sample greater than the first traffic threshold; α is the proportion of traffic greater than the first traffic threshold, α = the number of traffic samples greater than the first traffic threshold / the total number of samples; t ∈ Ω upper is the traffic value at time t greater than the corresponding first traffic threshold; Ωlower is a set of time instants in the flow sample that are less than or equal to the first flow threshold value, t∈Ω lower is a flow value at time t that is less than or equal to the corresponding first flow threshold value; W θ is all trainable parameters in the flow prediction model; v t+1 is a true value corresponding to the flow value at time t+1; is a prediction value of the flow prediction model; M is a length of a historical data window.

[0028] Embodiments of the present application also provide a communication network flow prediction device, comprising:

[0029] a first obtaining module, configured to obtain cell flow data;

[0030] a first determining module, configured to determine a first flow threshold value according to the cell flow data;

[0031] a second determining module, configured to determine an adjacency matrix in graph convolution of a flow prediction model;

[0032] a second obtaining module, configured to input the cell flow data, the first flow threshold value and the adjacency matrix into the flow prediction model to obtain a prediction result.

[0033] Embodiments of the present application also provide a communication network flow prediction device, comprising a transceiver and a processor;

[0034] the processor is configured to:

[0035] obtain cell flow data;

[0036] determine a first flow threshold value according to the cell flow data;

[0037] determine an adjacency matrix in graph convolution of a flow prediction model;

[0038] input the cell flow data, the first flow threshold value and the adjacency matrix into the flow prediction model to obtain a prediction result.

[0039] Embodiments of the present application also provide a communication network flow prediction device, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements steps of the communication network flow prediction method described above when executing the program.

[0040] Embodiments of the present application also provide a readable storage medium, having a computer program stored thereon, wherein the program implements steps of the communication network flow prediction method described above when executed by a processor.

[0041] The present application has the following beneficial effects:

[0042] The above scheme can improve the accuracy of traffic prediction by determining a first traffic threshold according to cell traffic data, determining an adjacency matrix, inputting the cell traffic data, the first traffic threshold and the adjacency matrix into a traffic prediction model, and obtaining a prediction result. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 is a flowchart of a communication network traffic prediction method according to an embodiment of the present application;

[0044] Figure 2 is a flowchart of a large traffic threshold calculation;

[0045] Figure 3 is a schematic diagram of a convolutional network structure;

[0046] Figure 4 is a module schematic diagram of a communication network traffic prediction device according to an embodiment of the present application;

[0047] Figure 5 is a structural diagram of a communication network traffic prediction device according to an embodiment of the present application. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0049] The terms "first", "second", and the like in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be exchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally a class, and are not limited to the number of objects, for example, the first object can be one or more. In addition, "and / or" in the specification and claims indicates at least one of the connected objects, and the character " / ", generally indicates that the front and rear associated objects are in an "or" relationship.

[0050] It is worth noting that the technology described in the embodiments of the present application is not limited to Long Term Evolution (LTE) / LTE-Advanced (LTE-A) systems, but can also be used in other wireless communication systems, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-carrier Frequency-Division Multiple Access (SC-FDMA) and other systems. The terms "system" and "network" in the embodiments of the present application are often used interchangeably, and the described technology can be used in the above-mentioned systems and radio technologies, as well as other systems and radio technologies. The following description describes a New Radio (NR) system for example purposes, and NR terminology is used in most of the following description, but these technologies can also be applied to applications other than NR system applications, such as 6th Generation (6G) communication systems.

[0051] The communication network traffic prediction method and device provided by the embodiments of the present application will be described in detail below in combination with the drawings, specific embodiments and application scenarios.

[0052] As shown in the Figure 1 At least one embodiment of the present application provides a communication network traffic prediction method, comprising:

[0053] Step 101, acquiring cell traffic data;

[0054] Step 102, determining a first traffic threshold according to the cell traffic data;

[0055] Step 103, determining an adjacency matrix in graph convolution of a traffic prediction model;

[0056] Step 104, inputting the cell traffic data, the first traffic threshold and the adjacency matrix into the traffic prediction model to obtain a prediction result.

[0057] It should be noted that the embodiment of the present application determines the first traffic threshold value according to the cell traffic data, and determines the adjacency matrix, and then inputs the cell traffic data, the first traffic threshold value and the adjacency matrix into the traffic prediction model to obtain a prediction result, so as to improve the accuracy of traffic prediction.

[0058] Optionally, in another embodiment of the present application, the further implementation process of step 101 is:

[0059] Obtaining initial traffic data, which refers to the original data obtained by the data acquisition device collecting the cell traffic.

[0060] Pretreating the initial traffic data to determine the cell traffic data.

[0061] It should be noted that the main purpose of processing the initial traffic data is to ensure that the data input into the traffic prediction model meets the model usage requirements, that is, the cell traffic data is data that meets the usage requirements of the traffic prediction model.

[0062] It should be noted that the pretreatment method in another embodiment of the present application can include at least one of the following:

[0063] A1, performing format conversion on the initial traffic data: the traffic data needs to be numerical data such as int or float.

[0064] A2, checking the missing situation of the time column of the initial traffic data, and completing the time points according to the time granularity.

[0065] A3, linearly filling the missing values of the initial traffic data.

[0066] Further, in another embodiment of the present application, the optional implementation of step 102 is:

[0067] Step 1021, grouping the cell traffic data according to time points;

[0068] It should be noted that the traffic data of different cells belonging to the same time point belongs to the same group.

[0069] Step 1022, determining the traffic value corresponding to the traffic data located at the preset percentage fractile point in each group as the first traffic threshold value.

[0070] It should be noted that the first traffic threshold value in the embodiment of the present application is used to screen the traffic data greater than the first traffic threshold value. Optionally, the first traffic threshold value can also be regarded as a large traffic threshold value, that is, the traffic data higher than the large traffic threshold value can be regarded as large traffic data.

[0071] The specific implementation process of step 102 is described in detail as follows.

[0072] In the operation and maintenance of a communication network, it is necessary to repair cells with large traffic first to maximize the reduction of loss. The traffic at each moment presents different rules, and similar trends may appear at the same moment every day. Therefore, the large traffic definition is proposed: the n% quantile point of N x M traffic data at the same moment of M cells in N historical days is defined as the large traffic threshold at this moment, and the traffic greater than the threshold at the same moment is called large traffic; where M is the number of cells in the communication network, N is the number of historical data days, and n can be determined according to the actual repair capacity for reference, and by default, 90% can be taken.

[0073] It should be noted that for obtaining the large traffic threshold, the traffic data belonging to a group can be sorted in descending order or ascending order, and the value of the traffic data sorted at a certain quantile point is selected as the large traffic threshold. For example, when the traffic data is sorted in ascending order, the data sorted at the 90% quantile point is selected as the large traffic threshold. Specifically, assuming that there are 100 traffic data in a group, each traffic data corresponds to a traffic value, and the traffic values are sorted in ascending order, when the data sorted at the 90% quantile point is selected as the large traffic threshold, the traffic value sorted at the 90th position is selected as the large traffic threshold.

[0074] For example, taking 15-minute time series data as an example, the large traffic threshold calculation process is as follows:

[0075] As shown in Figure 2 , the data of the last natural month in the historical data is selected, and then the data is divided into two batches according to the week (Monday to Friday) and the weekend (Saturday and Sunday), and then grouped according to the time (eg: 00:00:00, 00:15:00,....), each group is sorted from large to small, and the 90% quantile point in the group is calculated as the large traffic threshold at this moment. Finally, the large traffic threshold table obtained is shown in Table 1, which has a 15-minute granularity, and there are 96 time points in a day, and 96*2=192 large traffic thresholds are obtained.

[0076] Table 1 Large traffic threshold table

[0077] Time of day Weekday high traffic threshold Weekend high traffic threshold 00:00 15 21 00:15 16 20 00:30 24 23 …… 23:45 21 24

[0078] Alternatively, in another embodiment of the present application, the adjacency matrix in the graph convolution of the traffic prediction model is improved. Specifically, the adjacency matrix in the graph convolution of the traffic prediction model is determined based on the cell type and the switching frequency between cells.

[0079] That is, the prior information of cell type (the base station in the hospital, school, shopping mall, etc.) and the switching relationship between cells (i.e., the switching frequency between cells) in the network management system is introduced in this application to improve the adjacency matrix in graph convolution, so that the traffic prediction model can better learn and express the spatial information of traffic data.

[0080] Optionally, based on the cell type, the specific implementation of the adjacency matrix in the graph convolution of the traffic prediction model is:

[0081] According to the formula: determine the adjacency matrix in the graph convolution of the traffic prediction model;

[0082] wherein A is the adjacency matrix in the graph convolution of the traffic prediction model, A i,j is the value corresponding to the i-th row and j-th column of the adjacency matrix, dis(i,j) is the distance between cell i and cell j, d_thres is a pre-set neighbor distance threshold, Type i is the type of cell i, Type j is the type of cell j, and switch_freq(i,j) is the switching frequency between cells, count i,j is the number of switching between cell i and cell j, and total_time is the number of time points of data.

[0083] It should be further pointed out that in another embodiment of the present application, the loss function in the traffic prediction model is also improved (it should be noted that the loss function is used to evaluate the difference between the predicted value and the true value of the model, and is the objective function of the neural network optimization. The process of neural network training or optimization is to minimize the loss function. The role of the loss function is to calculate the gap between the forward calculation result of each iteration of the neural network and the true value, so as to guide the next training to the correct direction). Specifically, the first traffic threshold is used to determine the loss function in the traffic prediction model.

[0084] Optionally, in another embodiment of the present application, the loss function is obtained in the following manner:

[0085] According to the formula:

[0086] determine the loss function;

[0087] wherein L is the loss function; Ωupper is the set of time points greater than the first traffic threshold in the traffic sample; and a is the proportion of traffic greater than the first traffic threshold. is the loss function; Ωupper is the set of time points greater than the first traffic threshold in the traffic sample; and a is the proportion of traffic greater than the first traffic threshold. upper is the traffic value at time t greater than the corresponding first traffic threshold, i.e. indicates that all time points with traffic values greater than the first traffic threshold are summed up, L2 norm of the prediction minus the true value; Ωlower is a set of time instants in the traffic sample that are less than or equal to the first traffic threshold, Ω lower Ωlower is a set of time instants in the traffic sample that are less than or equal to the first traffic threshold, t e Ω lower is less than or equal to the corresponding first traffic threshold, i.e. denotes the sum of all traffic values that are less than or equal to the first traffic threshold. summed up; W θ is all trainable parameters in the traffic prediction model; v t+1 is the true value corresponding to the traffic value at time instant t+1. is the prediction value of the traffic prediction model; M is the length of the historical data window.

[0088] It should be noted that a = the number of traffic samples greater than the first traffic threshold / the total number of samples (i.e. the number of large traffic samples / the total number of samples, where the total number of samples is equal to the length of the historical data window M, and the number of large traffic samples is equal to the number of large traffic in the M samples), where the smaller a is, the smaller the proportion of large traffic samples, and the loss (Loss) given to the large traffic sample set in the present application is given a greater weight, so that the loss function focuses on the training of large traffic data.

[0089] Specifically, in the embodiments of the present application, a classic spatio-temporal graph convolution network (STGCN) is used as the basis, and the loss (Loss) function thereof is improved in view of the characteristics of communication traffic and the application scenarios of prediction results.

[0090] Specifically, GCN is used to extract spatial features, and time-domain CNN is used to extract time features, and the network structure is as follows:

[0091] The entire architecture is composed of three parts, from Figure 3 The left side of the figure shows that the entire architecture hierarchy is as follows: input -> ST-Conv Block (spatio-temporal graph convolution module) -> ST-Conv Block -> Output Block (output module) -> output. Figure 3 The middle of the figure shows the composition structure of the ST-Conv Block, and each ST-Conv Block is composed of two Gated Temporal Convolution layers (temporal gated convolution layers) sandwiching a Graph Convolution layer (graph convolution layer). Figure 3 The right side of the figure shows the structure of the temporal gated convolution, which is composed of a 1D convolution and a Gated Linear Unit (GLU).

[0092] Specifically, the further implementation process of step 104 of the embodiment of the application is as follows:

[0093] Step 1041, cell traffic data and an adjacency matrix are input into a convolutional network of a traffic prediction model (i.e., through a process of Figure 3 convolution), to obtain output data;

[0094] Step 1042, the convolutional output data, a true value, and a first traffic threshold are calculated by a loss function;

[0095] Step 1043, according to the derivative of the loss function, the error is returned along the direction of the smallest gradient to correct the weight values in the forward calculation formula;

[0096] Step 1044, the above steps are repeated until the loss function value reaches a satisfactory value, and the iteration is stopped, to obtain a trained model;

[0097] Step 1045, the trained model is used for prediction to obtain a prediction result, which is the traffic data at a future time after the input traffic data.

[0098] It should be noted that, in the embodiment of the application, the cell traffic data and the obtained adjacency matrix are input into a traffic prediction model improved by a loss function for training, the model hyperparameters (the hyperparameters refer to the parameters that can be adjusted in the model) are adjusted, and finally a trained traffic prediction model is obtained. The model is used to predict the traffic index, and the prediction result can support various network operation works.

[0099] In summary, at least one embodiment of the application proposes a definition of large traffic for the specific scenario of cell traffic in a communication network, constructs an adjacency matrix with more spatial characteristics, realizes the targeted iterative optimization of the prediction model, improves the data prediction effect, provides priority basis for cell fault repair and resource scheduling, more reasonably uses resources, and improves customer satisfaction. The application can achieve the following beneficial effects:

[0100] 1. Improve the accuracy of traffic data prediction;

[0101] 2. The spatial relationship between cells is fully expressed, and more effective data spatial characteristics are learned;

[0102] 3. The prediction model is tightly coupled with subsequent applications.

[0103] As shown in Figure 4 at least one embodiment of the application also provides a communication network traffic prediction device 400, which comprises:

[0104] A first acquisition module 401 is configured to acquire cell traffic data;

[0105] The first determining module 402 is configured to determine a first traffic threshold according to the cell traffic data.

[0106] The second determining module 403 is configured to determine an adjacency matrix in graph convolution of a traffic prediction model.

[0107] The second obtaining module 404 is configured to input the cell traffic data, the first traffic threshold and the adjacency matrix into the traffic prediction model to obtain a prediction result.

[0108] Optionally, the first determining module 402 comprises:

[0109] The grouping unit is configured to group the cell traffic data according to time points, wherein traffic data of different cells belonging to a same time point belongs to a same group.

[0110] The determining unit is configured to determine, as the first traffic threshold, a traffic value corresponding to traffic data located at a preset percentage fractile point in each group.

[0111] Optionally, the second determining module 403 is configured to:

[0112] determine the adjacency matrix in graph convolution of the traffic prediction model based on cell types and inter-cell switching frequency.

[0113] Optionally, the second determining module 403 is specifically configured to:

[0114] determine the adjacency matrix in graph convolution of the traffic prediction model according to a formula:

[0115] wherein A is the adjacency matrix in graph convolution of the traffic prediction model, A i,j is a value corresponding to an i-th row and a j-th column of the adjacency matrix, dis(i, j) is a distance between cell i and cell j, d_thres is a pre-set neighbor distance threshold, Type i is a type of cell i, Type j is a type of cell j, and switch_freq(i, j) is an inter-cell switching frequency, count i,j is a switching number of cell i and cell j, and total_time is a time point number of data.

[0116] Optionally, a loss function in the traffic prediction model is determined based on the first traffic threshold.

[0117] Optionally, the loss function is obtained in the following manner:

[0118] according to a formula:

[0119] ​ determine a loss function;

[0120] wherein, is the loss function; Ωupperis a set of time points in the traffic sample that are greater than a first traffic threshold; a is a proportion of traffic that is above the first traffic threshold, a = number of traffic samples that are greater than the first traffic threshold / total number of samples; t ∈ Ω upper is a traffic value at time t that is greater than the corresponding first traffic threshold; Ω lower is a set of time points in the traffic sample that are less than or equal to the first traffic threshold, t ∈ Ω lower is a traffic value at time t that is less than or equal to the corresponding first traffic threshold; W θ is all trainable parameters in the traffic prediction model; v t+1 is a true value corresponding to a traffic value at time t+1; is a predicted value of the traffic prediction model; M is a length of a historical data window.

[0121] It should be noted that the apparatus provided by at least one embodiment of the present application is an apparatus capable of executing the above-mentioned communication network traffic prediction method, and all embodiments of the above-mentioned communication network traffic prediction method are applicable to the apparatus and can achieve the same or similar beneficial effects.

[0122] At least one embodiment of the present application further provides a communication network traffic prediction apparatus, and the terminal includes a transceiver and a processor;

[0123] The processor is configured to:

[0124] obtain cell traffic data;

[0125] determine a first traffic threshold according to the cell traffic data;

[0126] determine an adjacency matrix in graph convolution of a traffic prediction model;

[0127] input the cell traffic data, the first traffic threshold, and the adjacency matrix into the traffic prediction model to obtain a prediction result.

[0128] Optionally, the processor is configured to:

[0129] group the cell traffic data according to time points, wherein traffic data of different cells belonging to the same time point belong to the same group;

[0130] determine a traffic value corresponding to traffic data located at a preset percentile point in each group as the first traffic threshold.

[0131] Optionally, the processor is configured to:

[0132] The adjacency matrix in the graph convolution of the traffic prediction model is determined based on a cell type and a switching frequency between cells.

[0133] Optionally, the processor is configured to:

[0134] According to the formula: determine the adjacency matrix in the graph convolution of the traffic prediction model;

[0135] wherein A is the adjacency matrix in the graph convolution of the traffic prediction model, A i,j is a value corresponding to an i-th row and a j-th column of the adjacency matrix, dis(i,j) is a distance between the cell i and the cell j, d_thres is a pre-set neighbor distance threshold, Type i is the type of the cell i, Type j is the type of the cell j, and switch_freq(i,j) is a switching frequency between cells, count i,j is a switching number of the cell i and the cell j, and total_time is a number of time points of data.

[0136] Optionally, a loss function in the traffic prediction model is determined based on a first traffic threshold.

[0137] Optionally, the processor is configured to:

[0138] According to the formula:

[0139] determine the loss function;

[0140] wherein is the loss function; Ωupper is a time point set greater than the first traffic threshold in the traffic sample; α is a proportion of traffic higher than the first traffic threshold, α = a number of traffic samples greater than the first traffic threshold / total number of samples; t ∈ Ω upper is a traffic value at the time point t greater than the corresponding first traffic threshold; Ω lower is a time point set less than or equal to the first traffic threshold in the traffic sample, t ∈ Ω lower is a traffic value at the time point t less than or equal to the corresponding first traffic threshold; W θ is all trainable parameters in the traffic prediction model; v t+1 is a true value corresponding to a traffic value at a time point t+1; is a prediction value of the traffic prediction model; and M is a length of a historical data window.

[0141] As Figure 5As shown, an embodiment of the present invention further provides a communication network traffic prediction device, including a processor 500, a transceiver 510, a memory 520, and a program stored in the memory 520 and executable on the processor 500; wherein the transceiver 510 is connected to the processor 500 and the memory 520 via a bus interface, wherein the processor 500 is configured to read the program in the memory and execute the following process:

[0142] Get cell traffic data;

[0143] Determining a first traffic threshold according to the cell traffic data;

[0144] Determine the adjacency matrix in the graph convolution of the traffic prediction model;

[0145] The cell traffic data, the first traffic threshold, and the adjacency matrix are input into the traffic prediction model to obtain a prediction result.

[0146] The transceiver 510 is configured to receive and send data under the control of the processor 500 .

[0147] Among them, Figure 5 In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically linking together various circuits of one or more processors represented by processor 500 and memory represented by memory 520. The bus architecture may also link together various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are all well known in the art and therefore will not be further described herein. The bus interface provides an interface. The transceiver 510 may be a plurality of components, i.e., a transmitter and a receiver, providing a unit for communicating with various other devices over a transmission medium, such as a wireless channel, a wired channel, an optical cable, or the like.

[0148] The processor 500 is responsible for managing the bus architecture and general processing, and the memory 520 can store data used by the processor 500 when performing operations.

[0149] Optionally, the processor 500 may be a CPU (central processing unit), an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array) or a CPLD (Complex Programmable Logic Device), and the processor may also adopt a multi-core architecture.

[0150] The processor calls a computer program stored in the memory to execute any of the methods provided by the embodiments of the application according to the executable instructions obtained. The processor and the memory can also be physically arranged separately.

[0151] Further, the processor 500 implements the following steps when executing the program:

[0152] The cell traffic data is grouped according to time, wherein the traffic data of different cells belonging to the same time belongs to the same group.

[0153] The traffic value corresponding to the traffic data located at the preset percentage quantile in each group is determined as the first traffic threshold.

[0154] Optionally, the processor 500 implements the following steps when executing the program:

[0155] Based on the cell type and the switching frequency between cells, the adjacency matrix in the graph convolution of the traffic prediction model is determined.

[0156] Optionally, the processor 500 implements the following steps when executing the program:

[0157] According to the formula: The adjacency matrix in the graph convolution of the traffic prediction model is determined;

[0158] Wherein, A is the adjacency matrix in the graph convolution of the traffic prediction model, A i,j is the value corresponding to the i-th row and j-th column of the adjacency matrix, dis(i,j) is the distance between cell i and cell j, d_thres is the pre-set neighbor distance threshold, Type i is the type of cell i, Type j is the type of cell j, and switch_freq(i,j) is the switching frequency between cells, count i,j is the number of switching between cell i and cell j, and total_time is the number of time of data.

[0159] Optionally, the loss function in the traffic prediction model is determined based on the first traffic threshold.

[0160] Optionally, the processor 500 implements the following steps when executing the program:

[0161] According to the formula:

[0162] The loss function is determined;

[0163] Wherein, is a loss function; Ωupperis a set of time points in the traffic sample that are greater than the first traffic threshold; a is a proportion of traffic that is above the first traffic threshold, a = number of traffic samples that are greater than the first traffic threshold / total number of samples; t e Ω upper is a traffic value at time t that is greater than the corresponding first traffic threshold; Ω lower is a set of time points in the traffic sample that are less than or equal to the first traffic threshold, t e Ω lower is a traffic value at time t that is less than or equal to the corresponding first traffic threshold; W θ is all trainable parameters in the traffic prediction model; v t+1 is a true value corresponding to the traffic value at time t+1; is a predicted value of the traffic prediction model; M is a length of a historical data window.

[0164] At least one embodiment of the present application also provides a communication network traffic prediction device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements each process in the communication network traffic prediction method embodiment and achieves the same technical effects, and details are not repeated here to avoid repetition.

[0165] At least one embodiment of the present application also provides a computer readable storage medium having a computer program stored thereon, wherein the program is executable on a processor to implement each process in the communication network traffic prediction method embodiment and achieve the same technical effects, and details are not repeated here to avoid repetition. The computer readable storage medium may be, for example, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, etc.

[0166] It should be noted that, in the present document, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the presence of additional identical elements in the process, method, article, or apparatus that comprises the element. Furthermore, it is to be understood that the method and apparatus of the present application can be carried out by more than one process, method, article, or apparatus either simultaneously, concurrently, or with intervening action that are carried out at the same time, either in a simultaneous fashion or in a fashion that is staggered in time. For example, the described methods can be carried out in a different order than described, and / or various steps can be combined or omitted, and / or additional steps can be added, without departing from the scope of the present application. Also, features described with respect to certain examples can be combined in other examples.

[0167] From the above description of the embodiments, it is apparent that the above-described method of the embodiments can be realized by means of software and general-purpose hardware platforms, and of course, can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such an understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as a ROM / RAM, a magnetic disk, or an optical disk) and includes a number of instructions for causing a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the methods described in the various embodiments of the present application.

[0168] The embodiments of the present application are described above with reference to the accompanying drawings, but the present application is not limited to the above-described specific embodiments, and the above-described specific embodiments are merely illustrative, rather than limiting, and a person of ordinary skill in the art can make many forms under the inspiration of the present application without departing from the scope of the present application and the scope protected by the claims.

Claims

1. A communication network traffic prediction method, characterized in that: include: Get cell traffic data; Determining a first traffic threshold according to the cell traffic data; Determine the adjacency matrix in the graph convolution of the traffic prediction model; Inputting the cell traffic data, the first traffic threshold, and the adjacency matrix into the traffic prediction model to obtain a prediction result; The determining of the adjacency matrix in the graph convolution of the traffic prediction model includes: Based on the cell type and the frequency of inter-cell handover, the adjacency matrix in the graph convolution of the traffic prediction model is determined.

2. The method according to claim 1, characterized in that The determining a first traffic threshold according to the cell traffic data includes: Grouping the cell traffic data according to time, wherein traffic data of different cells at the same time belong to the same group; The flow value corresponding to the flow data at the preset percentile in each group is determined as the first flow threshold.

3. The method according to claim 1, characterized in that The method of determining the adjacency matrix in the graph convolution of the traffic prediction model based on the cell type and the frequency of inter-cell handover includes: According to the formula: Determine the adjacency matrix in the graph convolution of the traffic prediction model; Among them, A is the adjacency matrix in the graph convolution of the traffic prediction model, A i,j is the value corresponding to row i and column j of the adjacency matrix, dis(i,j) is the distance between cells i and j, d_thres is the preset neighbor distance threshold, Type i is the type of cell i, Type j is the type of cell j, switch_freq(i,j) is the frequency of switching between cells, count i,j is the number of handovers between cell i and cell j, and total_time is the number of data moments.

4. The method according to claim 1, wherein The loss function in the traffic prediction model is determined based on a first traffic threshold.

5. The method according to claim 4, characterized in that The loss function is obtained as follows: According to the formula: Determine the loss function; in, is the loss function; Ω upper is the set of moments in the traffic sample that is greater than the first traffic threshold; α is the proportion of traffic that is higher than the first traffic threshold, α = the number of traffic samples greater than the first traffic threshold / the total number of samples; t∈Ω upper The flow rate value at time t is greater than the corresponding first flow rate threshold; Ω lower is the time set of traffic samples that is less than or equal to the first traffic threshold, t∈Ω lower The flow value at time t is less than or equal to the corresponding first flow threshold; W θ are all trainable parameters in the traffic prediction model; v t+1 is the true value corresponding to the flow value at time t+1; is the predicted value of the traffic prediction model; M is the length of the historical data window.

6. A communication network traffic prediction device, characterized in that: include: The first acquisition module is used to obtain cell traffic data; A first determining module, configured to determine a first traffic threshold according to the cell traffic data; A second determination module is used to determine the adjacency matrix in the graph convolution of the traffic prediction model; A second acquisition module is used to input the cell traffic data, the first traffic threshold and the adjacency matrix into the traffic prediction model to obtain a prediction result; The second determining module is specifically configured to: Based on the cell type and the frequency of inter-cell handover, the adjacency matrix in the graph convolution of the traffic prediction model is determined.

7. A communication network traffic prediction device, characterized in that: including a transceiver and a processor; The processor is configured to: Get cell traffic data; Determining a first traffic threshold according to the cell traffic data; Determine the adjacency matrix in the graph convolution of the traffic prediction model; Inputting the cell traffic data, the first traffic threshold, and the adjacency matrix into the traffic prediction model to obtain a prediction result; The determining of the adjacency matrix in the graph convolution of the traffic prediction model includes: Based on the cell type and the frequency of inter-cell handover, the adjacency matrix in the graph convolution of the traffic prediction model is determined.

8. A communication network traffic prediction device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the communication network traffic prediction method according to any one of claims 1 to 5 are implemented.

9. A readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the communication network traffic prediction method according to any one of claims 1 to 5 are implemented.

Images